Clustering individuals’ temporal patterns of affective states, hunger, and food craving by latent class vector-autoregression

Abstract

Background Eating plays an important role in mental and physical health and is influenced by affective (e.g., emotions, stress) and appetitive (i.e., food craving, hunger) states, among others. Yet, substantial temporal variability and marked individual differences in these relationships have been reported. Exploratory data analytical approaches that account for variability between and within individuals might benefit respective theory development and subsequent confirmatory studies. Methods Across 2 weeks, 115 individuals (83% female) reported on momentary affective states, hunger, and food craving six times a day. Based on these ecological momentary assessment (EMA) data we investigated whether latent class vector-autoregression (LCVAR) can identify different clusters of participants based on similarities in their temporal associations between these states. Results LCVAR allocated participants into three distinct clusters. Within clusters, we found both positive and negative associations between affective states and hunger/food craving, which further varied temporally across lags. Associations between hunger/food craving and subsequent affective states were more pronounced than vice versa. Clusters differed on eating-related traits such as stress-eating and food craving as well as on EMA completion rates. Discussion LCVAR provides novel opportunities to analyse time-series data in affective science and eating behaviour research and uncovers that traditional models of affect-eating relationships might be overly simplistic. Temporal associations differ between subgroups of individuals with specific links to eating-related traits. Moreover, even within subgroups, differences in associations across time and specific affective states can be observed. To account for this high degree of variability, future research and theories should consider individual differences in direction and time lag of associations between affective states and eating behaviour, daytime and specific affective states. In addition to that, methodological implications for EMA research are discussed.

Goal pursuit increases more after dietary success than after dietary failure: examining conflicting theories of self-regulation using ecological momentary assessment

“…All items were answered on a horizontal slider ranging from 0 (= ‘very little’/ ‘not at all’) to 100 (= ‘very much’/ ‘very’). EMA-items not included in the current analyses such as craving, hunger or affect are described elsewhere [ 29 , 30 ].…”

Predicting food craving in everyday life through smartphone-derived sensor and usage data

“…Participants included in the present secondary data analysis came from a larger randomized control trial (DRKS, DRKS00017493). 1 Other parts of the data have previously been used in the studies by Pannicke et al ( 19 ) and Kaiser and Butter ( 17 ). Participants were recruited to be motivated to maintain or reduce their body weight.…”

Toward Individualized Prediction of Binge-Eating Episodes Based on Ecological Momentary Assessment Data: Item Development and Pilot Study in Patients With Bulimia Nervosa and Binge-Eating Disorder

“…Thus, the present results are highly specific, for example, to the used "prediction interval" of 2.5 hours between predictors and outcome. This could be problematic as it has been shown that emotions and eating can influence each other at different time intervals [60].…”